Ferri, M.; Mangili, F.; Viano, G., 1993. Projective pose estimation of linear and quadratic primitives in monocular computer vision. CVGIP: Image understanding Vol.58, No.1, July, pp. 66-84  Home/Search   Document Not in Database   Summary   Related Articles   Check

....noise usually average out between features and because of the redundancy of information brought by the larger number of features. We have implemented two methods for estimating pose in our system: the first is closed form and based on the perspective inversion of four coplanar feature points [11] and the second is iterative and takes as input a set of non coplanar points [8] Though looking for coplanar features seemed a natural way to go in an urban environment, computations with such features have however turned out to be unstable and thus difficult to use in our application. We will ....

....Euler angles) and the three parameters of the translation matrix T . 3. 2 The Method of Ferri et al. The first method we have implemented for estimating pose is the exact perspective inversion of four coplanar points, which follows from the invariance of the cross ratio of four collinear points [11]. We thoroughly tested this method of determining viewpoint, but it turned out to be highly unreliable in our experiments: 5 Figure 4: Viewpoint determination with Ferri s method given four perfectly coplanar points on a plane of a calibration pattern and their 2D correspondents. When applied ....

[Article contains additional citation context not shown here]

Ferri, M., Mangili, F., and Viano, G. (1993). Projective pose estimation of linear and quadratic primitives in monocular computer vision. CVGIP: Image Understanding, 58(1):66--84.

....in [48] and this method, the real object tests show average orientation angle errors of 0.80 degrees or less and average positional errors of 1.28 mm or less with the largest error in the depth axis. Ferri et al. examine pose estimation through perspective inversion for four di#erent problems [50]. These problems are: 1) four coplanar line segments; 2) three orthogonal line segments; 3) a circular arc; and (4) a quadric of revolution. For case two, the method gives a second degree equation as the solution. Case three for elliptical images finds the orientation of planes that intersect ....

....magnitudes of the # m left and right vectors are still unknown. These can be expressed in terms of the transformation between the left and right reference frames, k left # mn left = k right # mn right ( # t # l) 5. 13) 150 where in this case # t is the translation and is equal to [50 0 0] T . An assumption is also implicitly made that no rotation occurs, so that the direction vector is the same for both the left and the right sides. Since only x translation is involved, the y and z components of the above equation can be used to solve for the scale factor k right : k right ....

M. Ferri, F. Mangili, and G. Viano, "Projective pose estimation of linear and quadratic primitives in monocular computer vision," CVGIP: Image Understanding, vol. 58, no. 1, pp. 66--84, July 1993.

.... and Davis [Dementhon 95] applied on a set of points (this method does not require any initial estimate) 4 Pose computation from various features Most pose estimation algorithms use simple features: points [Haralick 89, Dementhon 95] lines [Dhome 89, Shakunaga 93, Kumar 94] or circles [Ferri 93] But only few papers have been devoted to the 2D 3D registration of curves. Kriegman [Kriegman 90] proposed an algebraic method to compute the pose of a curved object from the observation of its occluding contours (it can be easily applied to perform 2D 3D registration) Unfortunately, this ....

M. Ferri, F. Mangili and G. Viano. Projective Pose Estimation of Linear and Quadratic Primitives in Monocular Computer Vision. CVGIP: Image Understanding, 58(1):66--84, July 1993.

....to be categorized as outlier or not [10] When curved features are considered, the problem is not so simple, as some parts of the 2D curves can perfectly match the 3D model whereas other parts can be erroneously matched. While numerous papers are dedicated to pose estimation from points or lines [4, 9, 14], only few papers have been devoted to the 3D 2D registration of curves [8, 13] However, these papers are not concerned with possible matching errors. The details of our robust pose computation algorithm (RPC) are given in this section. Emphasis is put on the robustness of the computation. First ....

M. Ferri, F. Mangili, and G. Viano. Projective Pose Estimation of Linear and Quadratic Primitives in Monocular Computer Vision. CVGIP: Image Understanding, 58(1):66-84, July 1993.

....image noise usually average out between features and because of the redundancy of information brought by the larger number of features. We have implemented two methods for estimating pose in our system: the rst is closed form and based on the perspective inversion of four coplanar feature points (Ferri et al. 1993) and the second is iterative and takes as input a set of non coplanar points (DeMenthon and Davis, 1995) Though looking for coplanar features seemed a natural way to go in an urban environment, computations with such features 8 M. O. Berger et al. have however turned out to be unstable and thus ....

....the Euler angles) and the three parameters of the translation matrix T . 3. 2 The Method of Ferri et al. The rst method we have implemented for estimating pose is the exact perspective inversion of four coplanar points, which follows from the invariance of the cross ratio of four collinear points (Ferri et al. 1993). We thoroughly tested this method of determining viewpoint, but it turned out to be highly unreliable in our experiments: When applied to a perfectly modelled calibration pattern as shown in Fig. 4, the exact perspective inversion method gives a viewpoint for which the four coplanar points used ....

[Article contains additional citation context not shown here]

Ferri, M., Mangili, F., and Viano, G. (1993). Projective pose estimation of linear and quadratic primitives in monocular computer vision. Cvgip: Image Understanding, 58(1):66-84.

....information to impose correspondence conditions, which is very attractive for applications. Several authors have remarked the importance of conics as basic image features and developed procedures for pose estimation, stereo and motion based on conics, for instance [1] 2] 3] 4] 5] 6] [7], 8] However, there are fewer articles dealing with conics than those devoted to Long Quan is with LIFIA CNRS INRIA, 46, avenue Felix Viallet, 38031 Grenoble, France. E mail: Long.Quan imag.fr. points and lines. In this paper, we are interested in the problem of conic correspondences and that ....

M. Ferri, F. Mangili, and G. Viano, "Projective pose estimation of linear and quadratic primitives in monocular computer vision", Computer Vision, Graphics and Image Processing, vol. 58, no. 1, pp. 66--84, July 1993.

....absolute invariant using the formule (2) 4. 5 Related work on conic reconstruction It is also important to note that several authors have remarked the importance of conics as basic image features and developed some procedures for pose estimation, stereo and motion based on conics, for instance [18, 2, 20, 32, 26, 37, 9, 22]. The conic reconstruction algorithm proposed in this paper is related to but different from those of Ma et. al [21, 20] and Safaee Rad et. al [32] They both have been interested in the Euclidean reconstruction of space conic and proposed different solutions to the problem. Ma et al. in [21, 20] ....

M. Ferri, F. Mangili, and G. Viano. Projective pose estimation of linear and quadratic primitives in monocular computer vision. Computer Vision, Graphics and Image Processing, 58(1):66--84, July 1993.

....parameters. 3.1. Initialization 3.1. 1 Perspective Inversion for Four Coplanar Points We first present the problem of computing the viewpoint from the observation of coplanar points (or segments) The technique we use gives a unique solution to the perspective inversion of four coplanar points [4]. The camera model assumed here is the pinhole model, with the image plane Omega situated at y = f (f is the focal length) The setup is as described in Figure 2, with Sigma the reference plane (the plane where the coplanar points lie) which we assume does not contain the origin O of the ....

M. Ferri, F. Mangili, and G. Viano. Projective Pose Estimation of Linear and Quadratic Primitives in Monocular Computer Vision. CVGIP: IU, 58(1):66--84, 1993.

No context found.

Ferri, M.; Mangili, F.; Viano, G., 1993. Projective pose estimation of linear and quadratic primitives in monocular computer vision. CVGIP: Image understanding Vol.58, No.1, July, pp. 66-84

No context found.

M. Ferri, F. Magnilli, G. Viano. Projective pose estimation of linear and quadratic primitives in monocular computer vision. Image understanding, Vol.58, No.1, pp.66-84, 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC